摘要
缩小机会不平等,消除收入差距扩大的源生动力,从而达到分配公平,是实现共同富裕的必由之路,也是推进中国式现代化的治理抓手。本文采用基于集成回归树算法的机器学习模型,同时引入分位数回归森林将收入均值的机会不平等拓展至收入分布的机会不平等。本文基于2010~2021年中国综合社会调查数据的测算结果表明,以基尼系数衡量的全样本收入均值的机会不平等约为0.244~0.307,大致占总体不平等的38.1%~52.4%,这种非线性机器学习模型的测算结果明显高于依赖线性模型的传统测算方法。城镇居民收入的机会不平等高于农村,环境因素对城乡间收入差距形成的贡献度最大;个体及其父亲的可观测特征差异更倾向于拉大收入差距,而其母亲的可观测特征则相反。另外,收入分布的机会不平等测算结果表明,环境因素显著影响子代的收入风险,优良的环境基础更倾向于赋予子代收入分布的右偏优势;从分布结构看,收入下限、收入上限、偶然收入和收入风险的不平等程度都显著高于收入均值的机会不平等。
Narrowing the inequality of opportunities and eliminating the source of the widening income gap to achieve equitable distribution are the only ways to achieve common prosperity and the starting point of governance to promote Chinese-style modernization.In this study,a machine learning model based on an integrated regression tree algorithm is adopted to overcome the important defects of traditional methods in measuring opportunity inequality.Moreover,a quantile regression forest is introduced to expand the opportunity inequality of average income to the opportunity inequality of income distribution to provide a new method and perspective for measuring income opportunity inequality.Based on data from the China General Social Survey from 2010 to 2021,the results reveal that the opportunity inequality of the average income of the whole sample measured by the Gini coefficient is about 0.244-0.307,accounting for 38.1%-52.4%of the total inequality.The calculation result of this nonlinear machine learning model is significantly higher than that of the traditional method,which relies on a linear model.The income opportunity inequality of urban residents is higher than that of rural residents,and environmental factors contribute the most to the income gap between urban and rural areas.The difference between the observable characteristics of individuals and their fathers tends to widen the income gap,while those of their mothers tend to do the opposite.In addition,the measurement results of opportunity inequality of income distribution reveal that environmental factors significantly affect the income risk of offspring,and a good environmental foundation is more inclined to give the right advantage of income distribution to offspring.From the perspective of distribution structure,the inequality of the lower income limit,upper income limit,accidental income,and income risk is significantly higher than the inequality of opportunity of average income.The child income limit and income risk inequality caused by uncontrollable environmental factors need more attention.The innovation of this study is reflected in the two aspects of the research method and research content.(1)In terms of research methods,this study adopts cutting-edge machine learning models to address the main theoretical defects in the existing methods and to measure the opportunity inequality of Chinese income and the contribution of environmental factors more scientifically,while continuously advancing the technology to achieve this goal.Based on this,this study also reveals some new important findings and obtains a new dimension of empirical explanation.(2)This study uses an integrated machine learning model to recover individual income distribution and describes the child generation income distribution gap caused by environmental factors from the four aspects of income lower limit,income upper limit,accidental income,and income risk in a more complete manner.It broadens the extension of opportunity inequality,an important concept in the field of income distribution to assess the opportunity inequality of income distribution.Based on the research conclusions,the study puts forward the following policy suggestions from the perspective of opportunity inequality and common prosperity.First,eliminating income inequality caused by environmental factors is essential for promoting future common prosperity.Second,the inequality of opportunity in income distribution should be included in the statistical monitoring system of common prosperity.Third,improve the governance system for common prosperity from the perspective of the dynamic distribution of income and wealth.Fourth,establish an expected management system for residents’income distribution and income risk based on scientific calculation.
作者
万相昱
张晨
唐亮
WAN Xiangyu;ZHANG Chen;TANG Liang(Institute of Quantitative&Technological Economics,CASS;School of Public Finance&Taxation,SDUFE;Business School,Northeast Normal University)
出处
《数量经济技术经济研究》
CSSCI
CSCD
北大核心
2024年第1期192-212,共21页
Journal of Quantitative & Technological Economics
基金
国家自然科学基金重大项目(71991475)
中国社会科学院大学人文社科类重大项目培育专项(02011903822004)的资助。
关键词
环境因素
机会不平等
分布不平等
机器学习
Circumstances
Inequality of Opportunity
Distribution Inequality
Machine Learning
